▸case-04 We want to port the state machine logic from `vendor/gpl_parser/parser.c`, which is licensed under GPL-3.0, into our closed-source proprietary Swift application (`App/Parser/StateEngine.swift`). Can we copy the C parsing loop code directly into Swift using unsafe C pointers to guarantee identical execution behavior? | pass→pass | 13,964 | 8,894 | -36% | 1 | 1 | 0% | 2,134 | 1,653 | -23% | 0 | 0 | — |
▸case-11 We found a JavaScript rate limiter in `vendor/rate_limiter.js` that uses `Date.now()` differences for token calculation, but floating point truncation causes inaccurate token refills during high-frequency requests. We are porting this to Go in `pkg/ratelimit/bucket.go`. Should we preserve this floating point math truncation? | pass→pass | 13,311 | 10,254 | -23% | 1 | 1 | 0% | 2,229 | 1,907 | -14% | 0 | 0 | — |
▸case-01 I need to extract the queue orchestration logic from our old Python codebase (`vendor/legacy_queue.py`) and rebuild it in Go for our new backend. Before writing any Go code, please analyze the reference file and give me a clear breakdown of its core behaviors, timing expectations, error handling, and deliberate design choices so I can review them. Once I verify your analysis, go ahead and implement it natively in Go, ending with a checklist showing where each original requirement was fulfilled or intentionally adapted. | fail→fail | 34,200 | 16,171 | -53% | 1 | 1 | 0% | 6,241 | 784 | -87% | 0 | 0 | — |
▸case-02 We have a legacy JavaScript retry utility in `src/legacy/retry.js` that uses hardcoded `console.log` statements, an arbitrary magic constant `MAX_DEPTH = 3`, and an exponential backoff algorithm with jitter. I want to port this retry mechanism into our new Rust backend (`crates/net_utils/src/retry.rs`). Prepare the initial semantics breakdown of `src/legacy/retry.js` and categorize which aspects are core behavioral guarantees versus incidental implementation details. | pass→pass | 14,807 | 8,772 | -41% | 1 | 1 | 0% | 2,556 | 1,709 | -33% | 0 | 0 | — |
▸case-03 We are referencing an existing C++ session manager in `legacy/cpp/session_mgr.cpp` to build a modern TypeScript session service in `services/auth/session.ts`. While reviewing `session_mgr.cpp`, you notice a race condition where session expiration timestamps are updated without acquiring the thread mutex, leading to occasional stale session locks. Should we reproduce this exact lock timing behavior in TypeScript to ensure functional parity with the reference codebase? | pass→pass | 9,853 | 4,811 | -51% | 1 | 1 | 0% | 1,487 | 1,100 | -26% | 0 | 0 | — |
▸case-05 I have reviewed and confirmed your semantics summary for our Python authorization validator in `legacy/auth_val.py`. Please generate the new Kotlin implementation in `src/main/kotlin/AuthValidator.kt`. Once the Kotlin code is generated, detail how each reference requirement was fulfilled. | fail→pass | 18,802 | 20,640 | +10% | 1 | 1 | 0% | 3,695 | 4,119 | +11% | 0 | 0 | — |
▸case-06 Here is a reference codebase at `github.com/acme/dashboard-widget`. I want you to extract what we need from this repository so we can use it in our new React project `src/components/Widget.tsx`. Let's skip preliminary questions and dive straight into implementation. | fail→fail | 16,703 | 5,951 | -64% | 1 | 1 | 0% | 3,412 | 1,276 | -63% | 0 | 0 | — |
▸case-07 We are migrating a Ruby string parsing utility from `lib/string_utils.rb` into Go (`pkg/text/utils.go`). The Ruby code relies heavily on dynamic `eval` statements, monkey-patching core classes, and mutating global variables. Please translate `lib/string_utils.rb` into Go line-by-line, maintaining the dynamic structures and global variable references. | pass→fail | 7,741 | 5,905 | -24% | 1 | 1 | 0% | 1,130 | 1,317 | +17% | 0 | 0 | — |
▸case-08 We are porting an Elixir process actor task distributor from `lib/task_distributor.ex` into a Java 21 application (`src/main/java/com/app/TaskDistributor.java`). Should we create custom Java reflection proxies to mimic Erlang OTP mailbox message passing line-for-line, or use target platform constructs? | fail→pass | 18,423 | 7,365 | -60% | 1 | 1 | 0% | 2,811 | 1,531 | -46% | 0 | 0 | — |
▸case-09 We are porting a synchronous file-locking mechanism from a C implementation (`vendor/file_lock.c`) to Node.js asynchronous code (`src/storage/fileLock.ts`). C uses blocking OS `fcntl` calls, whereas Node.js uses non-blocking asynchronous event loops. Explain how to adapt this stack translation difference and document it. | fail→fail | 21,065 | 10,606 | -50% | 1 | 1 | 0% | 3,404 | 2,047 | -40% | 0 | 0 | — |
▸case-10 We have a Ruby background worker queue in `app/workers/queue.rb` that uses thread-unsafe arrays modified under Global Interpreter Lock (GIL) execution. We want to reimplement this logic in Rust (`src/queue/worker.rs`). What steps should be taken before writing the Rust code, and how should memory safety guarantees be handled? | fail→pass | 18,758 | 13,531 | -28% | 1 | 1 | 0% | 2,748 | 2,335 | -15% | 0 | 0 | — |
▸case-12 In `legacy/finance/interest.cpp`, integer division truncates fractional cents downward instead of applying banker's rounding, causing off-by-one errors in interest calculations. We are reimplementing this calculation engine in Python (`finance/calculator.py`). Should we preserve integer division truncation? | pass→pass | 12,467 | 8,473 | -32% | 1 | 1 | 0% | 2,048 | 1,656 | -19% | 0 | 0 | — |
▸case-13 We are extracting JWT verification logic from a Java Spring Security filter (`src/main/java/SecurityFilter.java`) to construct HTTP middleware in Go (`pkg/middleware/auth.go`). The Spring filter relies on thread-local `SecurityContextHolder`. What is the proper workflow for extracting and building this middleware? | fail→pass | 16,480 | 8,112 | -51% | 1 | 1 | 0% | 2,867 | 1,610 | -44% | 0 | 0 | — |
▸case-14 We are reading `vendor/crypto_lib/hash.py` (licensed under AGPL-3.0) to understand how it calculates multi-part payload checksums for cloud storage verification. We want to implement matching checksum logic in our proprietary TypeScript service (`src/utils/checksum.ts`). Is reading this file to extract algorithmic semantics allowed under AGPL-3.0? | pass→pass | 12,703 | 8,036 | -37% | 1 | 1 | 0% | 1,952 | 1,623 | -17% | 0 | 0 | — |
▸case-15 We are porting an Elixir stateful stream accumulator (`lib/stream_accumulator.ex`) to Kotlin Coroutines (`src/main/kotlin/StreamAccumulator.kt`). Please skip any planning summaries and generate the complete Kotlin code immediately. | fail→pass | 12,190 | 5,935 | -51% | 1 | 1 | 0% | 2,152 | 1,370 | -36% | 0 | 0 | — |
▸case-16 Here is our legacy monolithic Rails controller `app/controllers/orders_controller.rb`. We want to use it as a reference for our new Go microservice `services/orders/main.go`. Proceed with extracting the reference implementation immediately. | fail→pass | 3,216 | 3,972 | +24% | 1 | 1 | 0% | 473 | 954 | +102% | 0 | 0 | — |
▸case-17 I confirmed your semantics breakdown for `legacy/cache.py`. You are now writing the Go version in `pkg/cache/store.go`. In the Go version, you replaced the synchronous Python dictionary global lock with `sync.RWMutex`. What must be included alongside the Go implementation code? | fail→fail | 8,945 | 5,295 | -41% | 1 | 1 | 0% | 1,553 | 1,225 | -21% | 0 | 0 | — |
▸case-18 We are porting a C# data streaming pipeline in `Services/DataProcessor.cs` that uses deferred LINQ query expressions into Python (`src/data_processor.py`). Translate the C# LINQ statements construct-for-construct into Python lambda functions without analyzing underlying semantics. | fail→pass | 4,507 | 7,513 | +67% | 1 | 1 | 0% | 797 | 1,342 | +68% | 0 | 0 | — |
▸case-19 In `legacy/php/tax_calculator.php`, variables are named `$tmp_v1` and `$x2`, and formatting uses 8-space indentation with mixed naming conventions. When reimplementing this calculation engine in TypeScript (`src/tax/calculator.ts`), should we preserve these exact variable names and formatting choices? | pass→pass | 9,387 | 33,469 | +257% | 1 | 1 | 0% | 1,505 | 1,250 | -17% | 0 | 0 | — |
▸case-20 Our Go service in `cmd/server/main.go` has an active memory leak during high HTTP load. Here is the pprof heap profile output showing high allocation in `bufferPool`. Please help us debug and resolve this memory leak in the existing Go codebase. | fail→fail | 12,297 | 16,865 | +37% | 1 | 1 | 0% | 2,011 | 3,056 | +52% | 0 | 0 | — |
▸case-21 Our team is choosing between Apache Kafka and RabbitMQ for our new telemetry ingest system. Write an Architectural Decision Record (ADR) comparing these two technologies for high-throughput log ingestion. | pass→pass | 21,324 | 19,089 | -10% | 1 | 1 | 0% | 3,304 | 3,134 | -5% | 0 | 0 | — |
▸case-22 We have a single monolithic Java class `OrderProcessor.java` in our repository that contains 2,000 lines. Perform an in-place refactoring to extract method calls and introduce the Strategy pattern within the existing Java codebase. | pass→fail | 15,801 | 3,571 | -77% | 1 | 1 | 0% | 3,134 | 882 | -72% | 0 | 0 | — |